@dubstrata/mcp-server

🚀 Dubstrata MCP Server Proxy

An ultra-lightweight, zero-dependency Model Context Protocol (MCP) server that acts as a secure local client proxy for the Dubstrata Agentic CDN and Graph RAG engine.

Instead of compiling and hosting heavyweight Python environment stacks, relational databases, vector storage indices, or Redis cache layers locally on your machine, this proxy server runs instantly via standard I/O (stdio) and forwards surgical graph requests to the live production server at https://api.dubstrata.com using your DUBSTRATA_API_KEY.

📦 Features

• Zero Local Footprint: No heavy local setup. Just Node.js.
• Surgical Graph RAG Tools: Exposes 10 powerful tools to explore, query, scrape, audit, and analyze the self-healing knowledge mesh.
• Dynamic IDE Support: Perfectly compatible with Claude Desktop, Cursor IDE, Windsurf, and other enterprise AI coding environments.
• Centralized Telemetry & ROI Logs: Automatically registers agent queries, measures token compression ratios, tracks real-time cost savings, and interacts seamlessly with the Ed25519 wallet micropayments framework.

⚙️ Quick Start

1. Claude Desktop Setup

Add this to your claude_desktop_config.json file (typically located at %APPDATA%\Claude\claude_desktop_config.json on Windows or ~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "dubstrata": {
      "command": "npx",
      "args": ["-y", "@dubstrata/mcp-server"],
      "env": {
        "DUBSTRATA_API_KEY": "your_api_key_here"
      }
    }
  }
}

2. Cursor IDE Integration

• Open Cursor and head to Settings -> Features -> MCP.
• Click + Add New MCP Server.
• Fill in the following details:
  • Name: Dubstrata
  • Type: command
  • Command: npx -y @dubstrata/mcp-server
• Click Save.
• Note: Make sure DUBSTRATA_API_KEY is loaded into your local environment variables so the shell can pick it up.

🛠️ Exposed Surgical Graph Tools

Once configured, the AI agent will automatically detect and utilize the following 10 surgical tools:

• ingest_knowledge – Ingests knowledge into the Graph RAG. Scrapes web URLs or accepts raw text.
• query_graph – Performs dense Graph RAG searches via vector similarity and graph traversal. Automatically triggers self-reflective JIT fallbacks (Wikipedia/Deep Crawls) if the context is missing.
• check_source_trust – Analyzes historical trust rates, fallacies, and critical security warnings/prompt injection risks of web domains.
• get_all_facts – Pulls all active claims and verified facts known about an entity directly from ArcadeDB.
• get_historical_timeline – Acts as a time machine to fetch superseded/historical versions of claims for audit trails.
• get_interaction_hooks – Extracts actionable developer integration next-steps linked to a document source.
• find_conflicts – Groups facts by source URLs to isolate direct factual contradictions or conflicts.
• get_entities_by_category – Finds and drills down into entities by categories (e.g., PERSON, ORGANIZATION, TECH).
• submit_feedback – Submits authenticated score and reason feedback for logged transactions to tune consensus models.
• compile_intelligence_report – Synthesizes deep RAG contexts into formal, multi-section analytical intelligence reports.

🔒 Configuration

You can customize the proxy using the following environment variables:

Variable	Description	Default
DUBSTRATA_API_KEY	Required. Your authorization key to sign queries and billing usage.	None
DUBSTRATA_API_URL	Optional domain override (e.g., for local developer environments or staging tests).	https://api.dubstrata.com

📄 License

This project is licensed under the MIT License.